The Empire Strikes Bach

Algorithms and copyright laws are stealing music from all of us

Getty Images
September 8, 2018

Sony has claimed that it owns the copyright for the works of Johann Sebastian Bach—more than 1,100 of them.

Now, you might think the fact that Bach died in 1750 would put his music safely in the public domain, seeing as how it's 178 years out of copyright (under the American system of author's death plus 70 years). But there the story was, appearing in several news accounts this past week, all prompted by a Boing Boing report about how "you can't play Bach on Youtube" without getting served with a takedown notice. Even the jazz historian Ted Gioia, as sane a music critic as exists these days, was prompted to tweet "Sony says they own his compositions."

To dig a little into the report, however, is to see how much the Boing Boing headline oversold the story. And yet, it's also to see how much the news about the YouTube incident has significantly undersold the story—which is really about the two-front war being waged on musical content. On one side, we've seen the consolidation of the Internet revolution into a small handful of platforms, and those platforms have stripped down to fractions of pennies what artists make from their music: "Get your song played a million times and get less than $5," as David Crosby complained  this August about royalty payments from Apple Music, Spotify, Amazon, YouTube, and the like.

Meanwhile, from the other side, the largest platforms have been joined by musical publishers and recording companies in an effort to lock down content. The technique—on display in last week's Bach incident—involves using new computer algorithms and neural-net AI systems to exploit provisions of DMCA (the Digital Millennium Copyright Act of 1998, which governs content-providers online).

In the most recent incident, a classical pianist named James Rhodes recorded himself playing Bach in his living room and posted the performance on his YouTube channel. YouTube almost immediately removed the video, sending Rhodes a notice that Sony Music Global claimed the copyright of 47 seconds from his music. Thus the headlines that Sony Owns Bach! that filled the Internet—and thus the mockery Sony received.

As it happens, the company genuinely does hold the copyright for several major Bach recordings, a collection crowned by Glenn Gould's performances. The YouTube claim was not that Sony owned Bach's music in itself. Rather, YouTube conveyed Sony's claim that Rhodes had recycled portions of a particular performance of Bach from a Sony recording.

The fact that James Rhodes was actually playing should have been enough to halt any sane person from filing the complaint. But that's the real point of the story. No sane person was involved, because no actual person was involved. It all happened mechanically, from the application of the algorithms in Youtube's Content ID system. A crawling bot obtained a complex digital signature for the sound in Rhodes's YouTube posting. The system compared that signature to its database of registered recordings and found a partial match of 47 seconds. The system then automatically deleted the video and sent a "dispute claim" to Rhodes's YouTube channel. It was a process entirely lacking in human eyes or human ears. Human sanity, for that matter.

Rhodes's case is not unique. In 2017, YouTube blanked out 90 seconds in the middle of Ukrainian pianist Valentina Lisitsa's performance of Bach's Partita No. 2. The Content ID system said it matched too much of Glenn Gould's 1957 recording, with a copyright owned by Sony.

Meanwhile, the German music professor Ulrich Kaiser wanted to develop a YouTube channel with free performances for teaching classical music. The first posting "explained my project, while examples of the music played in the background. Less than three minutes after uploading, I received a notification that there was a Content ID claim against my video." So he opened a different YouTube account called "Labeltest" to explore why he was receiving claims against public-domain music. Notices from YouTube quickly arrived for works by Bartok, Schubert, Puccini, Wagner, and Beethoven. Typically, they read, "Copyrighted content was found in your video. The claimant allows its content to be used in your YouTube video. However, advertisements may be displayed."

And that "advertisements may be displayed" is the key. Professor Kaiser wanted an ad-free channel, but his attempts to take advantage of copyright-free music quickly found someone trying to impose advertising on him—and thereby to claim some of the small sums that advertising on a minor YouTube channel would generate.

Last January, an Australian music teacher named Sebastian Tomczak had a similar experience. He posted on YouTube a 10-hour recording of white noise as an experiment. "I was interested in listening to continuous sounds of various types, and how our perception of these kinds of sounds and our attention changes over longer periods," he wrote of his project. Most listeners would probably wonder how white noise, chaotic and random by its nature, could qualify as a copyrightable composition (and wonder as well how anyone could get through 10 hours of it). But within days, the upload had five different copyright claims filed against it. All five would allow continued use of the material, the notices explained, if Tomczak allowed the upload to be "monetized," meaning accompanied by advertisements from which the claimants would get a share.

All of this happens because of a basic imbalance formed by general American copyright law and exacerbated by the 1998 DMCA. Technically, what these monetizing copyright notices do is a crime. It is simple fraud, a state-law case, to gain money by false claims. It's also a violation of Section 506(c) of the Copyright Act of 1976, which makes fraudulent uses of copyright notices a crime. Case law (see, for example, the 2015 civil case Lenz v. Universal Music Corp.) has generally read DMCA to require copyright owners to consider in good faith the possibility of fair use before they issue a takedown notice.

But the chance is nonexistent that a prosecutor will take on a criminal case for what looks, in each instance, to be a minute amount of money. And civil cases founder on the lack of proof of bad faith in the computer-generated Content ID complaints, generated by an algorithm without any human interaction. Google, which owns YouTube and created the Content ID system, has an interest in making sure that no human being is involved in the initial process, since the lack of human judgment offers them a safe harbor.

In other words, it's easy to generate a complaint about the use of copyrighted material and hard to generate a complaint about the misuse of claims to copyright. And so we get headlines like "you can't play Bach on Youtube."

The quickest fix for the current absurdity would be a checkmark system akin to one already in place. If a YouTube poster accumulates too many copyright violations, that poster's account is shut down. In a parallel way, YouTube could make it so that if a content provider accumulated too many rejected copyright claims, their access to the Content ID system would be closed off.

Unfortunately, Content ID is a service sold to those content providers. The chance is around nil that Google would cut off Sony for copyright errors generated by Google's own system. That leaves us with the alternative of the long ugly slog of what every story like Sony Owns Bach! suggests is coming: the blunt instrument of a class-action suit, with people like the pianist James Rhodes as its principals.

Either that, or we do just give over ownership of Bach to Sony and Google.

Published under: Music Reviews , Technology